IT equipment maintenance task dynamic scheduling method fused with fault prediction

Abstract

The invention discloses an IT equipment maintenance task dynamic scheduling method fused with fault prediction, and relates to the technical field of maintenance task scheduling. The IT equipment maintenance task dynamic scheduling method fused with fault prediction comprises the following steps: constructing an equipment state spatial-temporal characteristic matrix and a historical fault frequency matrix of IT equipment, and determining an occurrence frequency matrix of each fault type of the IT equipment in combination with a working condition correction factor stored in a database; determining a fault maintenance task amount based on the number of IT devices and the occurrence frequency matrix of each fault type; according to the system and the method, the expected completion state of the fault maintenance task is determined based on the current working state of the resident maintenance personnel, and the working mode and personnel of the resident maintenance personnel are adjusted based on the expected completion state of the fault maintenance task, so that the fault quantity can be intelligently evaluated, and the manual workload distribution can be optimized to improve the fault repair efficiency. Downtime is reduced, and user experience is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of maintenance task scheduling, and in particular to a method for dynamic scheduling of IT equipment maintenance tasks integrating fault prediction. Background Art

[0002] With the widespread application of smart terminals in various industries, the diversification of their types and functions has significantly increased the frequency of hardware and software failures. In special fields such as hospitals that are open to the public and related to life and health, once a smart terminal fails, it will directly interfere with the smooth conduct of doctors' consultation, diagnosis and intervention work, thereby having a negative impact on the patient's treatment process; in government office scenarios, smart terminals are important tools for government affairs processing, information transmission, etc. When a failure occurs, it may cause government affairs delays, poor information flow, and other problems, seriously affecting the efficiency and quality of public services; in terms of enterprise operations, whether it is production management, marketing or customer service, smart terminals play an indispensable role. The occurrence of failures will disrupt business processes, reduce the company's operating efficiency, and increase operating costs. Traditional fault repair methods mainly rely on manual inspections and maintenance. This method is not only inefficient, but also difficult to accurately assess the matching relationship between the amount of faults and the amount of manual workload, resulting in uneven resource allocation, slow response speed, and affecting overall work efficiency. Therefore, there is an urgent need for a system and method that can intelligently assess the amount of faults and optimize the allocation of manual workload to improve fault repair efficiency, reduce downtime, and enhance user experience. Summary of the invention

[0003] To achieve the above objectives, the present invention is implemented through the following technical solutions: a dynamic scheduling method for IT equipment maintenance tasks integrating fault prediction, comprising the following steps: constructing a device status spatiotemporal feature matrix and a historical fault frequency matrix of IT equipment, and determining a frequency matrix of each fault type of IT equipment in combination with the operating condition correction factor stored in a database; determining the amount of fault maintenance tasks based on the number of IT equipment and the frequency matrix of each fault type; determining the expected completion status of the fault maintenance task based on the current working status of the on-site maintenance personnel, and adjusting the working mode and personnel of the on-site maintenance personnel based on the expected completion status of the fault maintenance task.

[0004] Further, constructing a spatio-temporal feature matrix of the device status of IT devices includes the following steps: performing device ID-timestamp correlation cleaning on historical fault data, extracting fault component codes and severity levels from maintenance logs through regular expressions, and generating a structured fault feature vector; calibrating the UTC time of device operation logs, aligning load curves with different sampling frequencies using the dynamic time warping algorithm, calculating the load change rate per hour through the first derivative, and generating an operation status time series matrix; constructing a joint interpolation space for temperature-humidity sensor and vibration spectrometer data, generating a minute-level synchronized data stream through cubic spline interpolation, and calculating the temperature-vibration covariance coefficient per device per hour to form an environmental impact tensor; obtaining the spatio-temporal feature matrix of the device status through spatio-temporal tensor product operation on the structured fault feature vector, the operation status time series matrix, and the environmental impact tensor; Constructing a historical fault frequency matrix includes the following steps: obtaining the device historical fault data from the database, grouping by device model, calculating the average annual incidence rate of each fault type, and constructing it into a historical fault frequency matrix.

[0005] Further, determining the occurrence frequency matrix of each fault type of IT devices includes the following steps: obtaining the trained multi-task prediction model stored in the database, the multi-task prediction model uses a dual-channel gated recurrent unit network, the main channel input is the spatio-temporal feature matrix of the device status , and the auxiliary channel input is the historical fault frequency matrix and the working condition correction factor ; Using the modified Poisson regression to output the occurrence frequency of each fault type of each IT device within the future time window; ; ; ; wherein, is the gating signal, is the gating weight matrix, is the Sigmoid function, is the calculation result of the fusion feature, is the element-wise product, is the bias term of the c-th fault type, is the sensitivity weight of the c-th fault type of the i-th IT device to the m-th dimension of the fusion feature, M is the total dimension number of the fusion feature, is the occurrence frequency of the c-th fault type of the i-th IT device within the future time window; According to the real-time data flow, dynamically adjust the predicted occurrence frequency, and construct it into the occurrence frequency matrix of each fault type of IT devices: , wherein, To dynamically adjust the occurrence frequency of predictions, which represents the frequency of the c-th type of failure occurring in the future time window for the i-th type of IT device, is the forgetting factor, is the length of the sliding window, is the actual number of secondary failures occurring within the length of the sliding window.

[0006] Furthermore, based on the number of IT devices and the failure occurrence frequency matrix for each type of failure, determine the amount of fault repair tasks. The process is as follows: , where is the amount of fault repair tasks, is the number of the i-th type of IT device.

[0007] Furthermore, based on the current working status of on-site maintenance personnel, determine the expected completion status of fault repair tasks, including the following steps: Obtain the maintenance work efficiency of each on-site maintenance personnel for each type of failure of each IT device, and calculate the average fault maintenance efficiency of each on-site maintenance personnel; Obtain the on-site time of each current on-site maintenance personnel, and calculate the manual maintenance workload: , where is the manual maintenance workload, is the average fault maintenance efficiency of the a-th on-site maintenance personnel, is the on-site time of the a-th on-site maintenance personnel, and A is the number of current on-site maintenance personnel; Subtract the manual maintenance workload from the amount of fault repair tasks to obtain the difference; If the difference is less than 0, the expected completion status of the fault repair task is unable to be completed; If the difference is not less than 0 and not greater than the set threshold, the expected completion status of the fault repair task is able to be completed; If the difference is greater than the set threshold, the expected completion status of the fault repair task is that there are too many on-site personnel.

[0008] Furthermore, based on the expected completion status of the fault repair task, adjust the working mode and personnel of the on-site maintenance personnel, including the following steps: If the expected completion status of the fault repair task is unable to be completed, conduct internal transfer of maintenance personnel; If the expected completion status of the fault repair task is able to be completed, maintain the current status of the on-site maintenance personnel; If the expected completion status of the fault repair task is that there are too many on-site personnel, conduct external transfer of maintenance personnel or change the working mode of the maintenance personnel.

[0009] Furthermore, conduct internal transfer of maintenance personnel, including the following steps: Determine the list of the most matching personnel from the list of schedulable personnel based on the fitness calculation model; Based on the constraint conditions, screen the personnel in the list of the most matching personnel to determine the best combination of internal transfer personnel.

[0010] Further, determining the most-matched personnel list from the list of schedulable personnel based on the fitness calculation model includes the following steps: determining the work characteristic matrix of each schedulable personnel in the list of schedulable personnel, including the set of historical maintenance fault types, the set of completion times of maintenance for each fault type, the set of completion durations of maintenance for each fault type, and the average work efficiency; inputting the work characteristic matrix of each schedulable personnel into the fitness calculation model to obtain the maintenance fitness of each schedulable personnel; arranging the maintenance fitness in descending order, and obtaining the top R schedulable personnel as the most-matched personnel list.

[0011] Further, the fitness calculation model is expressed as: ; Wherein, is the maintenance fitness, is the average work efficiency, is the set of historical maintenance fault types, is the set of completion times of maintenance for each fault type, is the set of completion durations of maintenance for each fault type, is the set of required historical maintenance fault types, is the set of required completion times of maintenance for each fault type, is the set of required completion durations of maintenance for each fault type, is the cosine similarity function.

[0012] Further, based on the constraint conditions, screening the personnel in the most-matched personnel list to determine the best internal transfer personnel combination, including the following steps: randomly combining the original on-site personnel with the personnel in the most-matched personnel list to obtain a maintenance personnel combination, where the maintenance personnel combination includes all the original on-site personnel and at least one personnel in the most-matched personnel list; evaluating each maintenance personnel combination based on the constraint conditions to determine the best maintenance personnel combination as the best internal transfer personnel combination, wherein the constraint conditions are: ; Wherein, is the fault repair task volume, is the average work efficiency of the p-th maintenance personnel in the maintenance personnel combination, is the on-site duration of the p-th maintenance personnel in the maintenance personnel combination, is the redundancy factor, is the operation and maintenance cost function.

[0013] The present invention has the following beneficial effects: The dynamic scheduling method for IT equipment maintenance tasks with integrated fault prediction constructs a spatio-temporal feature matrix of equipment status and a historical fault frequency matrix, and combines the working condition correction factors in the database to determine the occurrence frequency matrix of each fault type of the IT equipment, thereby accurately determining the amount of fault maintenance tasks. At the same time, based on the working status of on-site maintenance personnel, it determines the expected completion of fault maintenance tasks, and accordingly makes reasonable adjustments to the personnel working mode and personnel. It can not only improve the accuracy of fault prediction, realize the intelligent evaluation and dynamic scheduling of maintenance tasks, optimize the allocation of maintenance resources, improve maintenance efficiency, reduce equipment downtime, but also improve the overall work efficiency and user experience. It is especially suitable for key fields where intelligent terminals are widely used and have high requirements for fault maintenance, and can effectively solve problems such as low maintenance efficiency and uneven resource allocation caused by frequent faults of intelligent terminals in key fields (such as hospitals, governments, enterprises, etc.).

[0014] Of course, it is not necessary for any product implementing the present invention to achieve all the above-mentioned advantages simultaneously. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a flowchart of the dynamic scheduling method for IT equipment maintenance tasks with integrated fault prediction according to the present invention.

[0016] Figure 2 It is a schematic diagram of the composition of the operation and maintenance cost of the dynamic scheduling method for IT equipment maintenance tasks with integrated fault prediction according to the present invention.

[0017] Figure 3 It is a schematic diagram of the operation and maintenance cost measurement model of the dynamic scheduling method for IT equipment maintenance tasks with integrated fault prediction according to the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0018] Please refer to Figure 1 , the embodiments of the present invention provide a technical solution: a dynamic scheduling method for IT equipment maintenance tasks with integrated fault prediction, including the following steps: constructing a spatio-temporal feature matrix of equipment status and a historical fault frequency matrix of the IT equipment, and combining the working condition correction factors stored in the database to determine the occurrence frequency matrix of each fault type of the IT equipment.

[0019] Construct the spatio-temporal feature matrix of the device status of IT devices, including the following steps: perform device ID-timestamp correlation cleaning on historical fault data, extract the fault component codes and severity levels from the maintenance logs through regular expressions, and generate a structured fault feature vector containing [device ID, timestamp, fault type code, maintenance measure hash value]; it can normalize the messy historical fault data, extract key information and form a structured fault feature vector. This can improve the availability and accuracy of the data, providing a reliable basis for subsequent fault prediction and analysis. Through the correlation of device ID and timestamp, the fault history of each device can be better traced and analyzed, helping to discover the laws and trends of faults.

[0020] Calibrate the UTC time of the device operation logs, use the dynamic time warping algorithm to align the load curves with different sampling frequencies, calculate the hourly load change rate through the first derivative, and generate an operation status time series matrix containing [device ID, timestamp, operation duration, start-stop times, load change gradient]; it can unify and regularize the device operation logs in terms of time, enabling data with different sampling frequencies to be aligned, thus more accurately reflecting the changes in the device operation status. By calculating the load change rate, the load fluctuations during the device operation can be captured, which is of great significance for predicting device faults because abnormal load changes are often precursors of faults.

[0021] Construct a joint interpolation space for the data of the temperature and humidity sensor (once every 5 minutes) and the vibration spectrometer (once every hour), generate a minute-level synchronized data stream through cubic spline interpolation, calculate the temperature-vibration covariance coefficient per device per hour, and form an environmental impact tensor of [device ID, timestamp, temperature mean, vibration RMS, covariance coefficient]; comprehensively consider the influence of environmental factors (such as temperature and vibration) on device operation. By constructing a joint interpolation space and generating a synchronized data stream, the relationship between environmental factors such as temperature, humidity, and vibration and device faults can be analyzed more precisely. Abnormal changes in temperature and vibration may lead to a decline in device performance or faults, so this step helps to more comprehensively evaluate the impact of the device operation environment on its reliability.

[0022] Obtain the device status spatio-temporal feature matrix through spatio-temporal tensor product operation on the structured fault feature vector, the operation status time series matrix, and the environmental impact tensor; effectively fuse different types of feature data (fault features, operation status, and environmental impacts) to form a comprehensive device status spatio-temporal feature matrix. This fusion can more comprehensively describe the device status, provide richer information for the subsequent fault prediction model, and improve the accuracy and reliability of prediction. Through spatio-temporal tensor product operation, the change patterns of the device status in time and space can be captured, which plays a key role in accurately predicting the occurrence of faults.

[0023] Construct a historical failure frequency matrix, including the following steps: Obtain the device historical failure data from the database, group it by device model, calculate the annual average occurrence rate of each failure type, and construct it into a historical failure frequency matrix.

[0024] Determine the occurrence frequency matrix of each failure type of IT devices, including the following steps: Obtain the trained multi-task prediction model stored in the database. The multi-task prediction model uses a dual-channel gated recurrent unit network. The main channel input is the device status spatio-temporal feature matrix , and the auxiliary channel input is the historical failure frequency matrix and the working condition correction factor ; By using the pre-trained multi-task prediction model, it can handle multiple related tasks simultaneously (such as the prediction of different failure types), improving the efficiency and performance of the model. The dual-channel gated recurrent unit network can effectively process time series data, capture the changing trend of the device status over time, and consider the historical failure frequency and the working condition correction factor at the same time, making the prediction more accurate and comprehensive.

[0025] Use the modified Poisson regression to output the occurrence frequency of each failure type of each IT device within the future time window; ; ; ; where, is the gating signal, is the gating weight matrix, is the Sigmoid function, is the calculation result of the fused feature, is the element-wise product, is the bias term of the c-th failure type, is the sensitivity weight of the c-th failure type of the i-th IT device to the m-th dimensional fused feature, M is the total number of dimensions of the fused feature, is the occurrence frequency of the c-th failure type of the i-th IT device within the future time window; The modified Poisson regression can handle count data (such as the number of failures) and can predict the future failure occurrence frequency. In this way, it is possible to understand in advance the possible failure situations of each IT device, provide a basis for the subsequent maintenance task arrangement, contribute to realizing preventive maintenance, and reduce the device downtime.

[0026] Dynamically adjust the predicted occurrence frequency according to the real-time data flow, and construct the occurrence frequency matrix of each failure type of IT devices: , where, To dynamically adjust the predicted occurrence frequency, which represents the frequency of the c-th type of failure occurring in the future time window for the i-th type of IT device, is the forgetting factor, is the sliding window length, is the actual number of secondary failures occurring within the sliding window length. The predicted occurrence frequency is dynamically adjusted based on the latest real-time data, making the prediction results closer to the actual situation. By introducing the forgetting factor and the sliding window length, the influence of historical data and new data can be balanced to a certain extent, improving the accuracy and timeliness of the prediction. The constructed occurrence frequency matrix provides a basis for determining the amount of fault repair tasks in the follow-up.

[0027] Determine the amount of fault repair tasks based on the number of IT devices and the occurrence frequency matrix of each type of failure.

[0028] , where, is the amount of fault repair tasks, is the number of the i-th type of IT device. By comprehensively considering the device quantity and the failure occurrence frequency, the amount of fault repair tasks is accurately determined, providing a scientific basis for subsequent maintenance task scheduling and resource allocation. This helps to solve problems such as uneven resource allocation and low maintenance efficiency caused by the inability to accurately evaluate the matching relationship between the amount of faults and the manual workload in traditional maintenance methods, thereby improving the overall maintenance efficiency and resource utilization efficiency.

[0029] Determine the expected completion status of fault repair tasks based on the current working status of on-site maintenance personnel, and adjust the working mode and personnel of on-site maintenance personnel based on the expected completion status of fault repair tasks.

[0030] Obtain the maintenance work efficiency of each on-site maintenance personnel for each type of failure of each IT device, and calculate the average fault maintenance efficiency of each on-site maintenance personnel; quantitatively evaluate the work efficiency of each on-site maintenance personnel, providing data support for subsequent task assignment and personnel scheduling. By understanding the maintenance efficiency of different maintenance personnel for different devices and types of failures, the maintenance tasks can be arranged more reasonably, improving the efficiency of the overall maintenance work.

[0031] Obtain the on-site time of each current on-site maintenance personnel, and calculate the manual maintenance workload: , where, is the manual maintenance workload, is the average fault maintenance efficiency of the a-th on-site maintenance personnel, $t_a$ is the on-site time of the $a$-th on-site maintenance personnel, and $A$ is the number of current on-site maintenance personnel; the workload that a maintenance personnel can undertake is calculated based on their on-site time, so as to more accurately evaluate the total work ability of the current maintenance personnel. By combining the on-site time with the maintenance efficiency, the maximum workload of each maintenance personnel within a given time can be obtained, providing a basis for subsequent task allocation.

[0032] Subtract the manual maintenance workload from the fault repair task volume to obtain the difference. If the difference is less than 0, the expected completion status of the fault repair task is unable to be completed; if the difference is not less than 0 and not greater than the set threshold, the expected completion status of the fault repair task is able to be completed; if the difference is greater than the set threshold, the expected completion status of the fault repair task is that there are too many on-site personnel. Intuitively compare the gap between the work ability of the maintenance personnel and the actual repair task volume, so as to judge whether the current maintenance personnel can complete the task. By calculating the difference, it can be clearly known whether the task can be completed on time, or whether there is a shortage or surplus of personnel.

[0033] If the expected completion status of the fault repair task is unable to be completed, then conduct internal transfer of maintenance personnel; if the expected completion status of the fault repair task is able to be completed, maintain the current status of on-site maintenance personnel; if the expected completion status of the fault repair task is that there are too many on-site personnel, then conduct external transfer of maintenance personnel or change the work mode of maintenance personnel (adopt semi-on-site or call service methods). Specifically, if the difference is less than 0, it means that the work ability of the maintenance personnel is insufficient and unable to complete the task; if the difference is not less than 0 and not greater than the set threshold, it means that the task can be completed, but certain efforts may be required; if the difference is greater than the set threshold, it means that the work ability of the maintenance personnel is excessive and there may be redundant personnel.

[0034] Conducting internal transfer of maintenance personnel includes the following steps: determining the most matching personnel list from the list of schedulable personnel based on the adaptability calculation model; screening the personnel in the most matching personnel list based on the constraint conditions to determine the best combination of internal transfer personnel. The adaptability calculation model comprehensively considers factors such as the historical maintenance fault types, the number of maintenance completions, the maintenance completion duration, and the average work efficiency of the maintenance personnel, and can ensure that the selected personnel are the most matching with the current task in terms of skills and experience, improving the quality and efficiency of the repair work.

[0035] On the basis of the most matching personnel, further consider the actual constraint conditions (such as the fault repair task volume, the average work efficiency of the maintenance personnel, the on-site duration, and the operation and maintenance cost, etc.), so as to determine the optimal combination of internal transfer personnel. In this way, not only can the smooth completion of the repair task be guaranteed, but also the rationality and economy of the personnel combination can be ensured, avoiding waste of human resources and unnecessary cost expenditure.

[0036] Determine the list of the most suitable personnel from the list of schedulable personnel based on the fitness calculation model, including the following steps: Determine the work characteristic matrix of each schedulable personnel in the list of schedulable personnel, including the set of historical maintenance fault types, the set of the number of times each fault type is maintained, the set of the duration of each fault type maintained, and the average work efficiency; Input the work characteristic matrix of each schedulable personnel into the fitness calculation model to obtain the maintenance fitness of each schedulable personnel; Arrange the maintenance fitness in descending order, and obtain the top R schedulable personnel as the list of the most suitable personnel.

[0037] The fitness calculation model is expressed as: ; Wherein, is the maintenance fitness, is the average work efficiency, is the set of historical maintenance fault types, is the set of the number of times each fault type is maintained, is the set of the duration of each fault type maintained, is the set of required historical maintenance fault types, is the set of the number of times each required fault type is maintained, is the set of the duration of each required fault type maintained, is the cosine similarity function. Through these data, it is possible to understand in detail the maintenance experience, the number of completions, the completion duration, and the average work efficiency of each schedulable personnel for different fault types, etc., providing a rich data basis for the subsequent fitness calculation, and helping to more accurately evaluate the matching degree of each schedulable personnel with the current maintenance task. Using the cosine similarity function, it is possible to effectively measure the similarity between the historical work characteristics of the maintenance personnel and the requirements of the current maintenance task. The cosine similarity is not affected by the length of the feature vector, and can better reflect the directional relationship between features, so as to more accurately evaluate the matching degree of the maintenance personnel and the task. This helps to screen out the most suitable personnel for the current task among many schedulable personnel, improving the efficiency and quality of the maintenance work.

[0038] Based on the constraint conditions, screen the personnel in the list of the most suitable personnel to determine the best combination of internal transfer personnel, including the following steps: Randomly combine the original on-site personnel with the personnel in the list of the most suitable personnel to obtain a maintenance personnel combination, and the maintenance personnel combination includes all the original on-site personnel and at least one personnel in the list of the most suitable personnel; Evaluate each maintenance personnel combination based on the constraint conditions to determine the best maintenance personnel combination as the best combination of internal transfer personnel, wherein the constraint conditions are: ; Wherein, is the amount of fault repair tasks, is the average work efficiency of the p-th maintenance personnel in the maintenance personnel combination, is the on-site duration of the p-th maintenance personnel in the maintenance personnel combination, is the redundancy factor, is the operation and maintenance cost function. It can comprehensively evaluate the generated maintenance personnel combination according to actual constraints (such as the amount of fault repair tasks, the average work efficiency of maintenance personnel, on-site duration, and operation and maintenance costs, etc.). In this way, the best maintenance personnel combination that can minimize the operation and maintenance costs while meeting the task requirements can be screened out. This not only ensures the smooth completion of repair tasks, but also optimizes the utilization of human resources, avoiding personnel redundancy and cost waste.

[0039] It should be noted that the operation and maintenance cost function The measurement method is as Figure 2 and Figure 3 shown: The operation and maintenance costs include direct labor costs, direct non-labor costs, indirect labor costs, and indirect non-labor costs in operation and maintenance services.

[0040] Direct labor costs include human resource expenses such as salaries, bonuses, and benefits of the personnel providing operation and maintenance services. Among them, the personnel providing operation and maintenance services include those directly participating in the service process, such as service managers, engineers, quality assurance personnel, etc. For personnel not fully committed to this operation and maintenance service work, their human resource expenses are calculated according to the proportion of the workload in their total workload.

[0041] Direct non-labor costs include: a) Office expenses, that is, the administrative office expenses incurred by the operation and maintenance service provider for providing this operation and maintenance service, such as office supplies, communication, mailing, printing, meetings, etc.; b) Travel expenses, that is, the travel expenses incurred by the operation and maintenance service provider for providing this operation and maintenance service, such as transportation, accommodation, travel subsidies, etc.; c) Training expenses, that is, the expenses incurred by the operation and maintenance service provider for arranging special training for providing this operation and maintenance service; d) Business expenses, that is, the expenses incurred by the operation and maintenance service provider for the auxiliary activities required to complete this operation and maintenance service work, such as review fees, acceptance fees, etc.; e) Procurement expenses, that is, the expenses incurred by the operation and maintenance service provider for purchasing special assets or services for this operation and maintenance service, such as spare parts, operation and maintenance tools, technical cooperation fees, patent fees, venues, etc.; f) Lease expenses: namely, the lease expenses required by the operation and maintenance service provider for this operation and maintenance service work, such as office space lease expenses (due to the limited office conditions of the operation and maintenance service demander, it is unable to provide office space for operation and maintenance service personnel, but the timeliness of the service is required), equipment lease expenses (when dealing with professional problems, it is often necessary to temporarily lease professional equipment; regularly provide disaster recovery drills as required by the demander, etc.); g) Others, that is, the expenses that are not listed in a) to f) but are indeed incurred by the operation and maintenance service provider for this service.

[0042] Indirect labor costs refer to the allocation of human resource expenses of non-project team personnel used by the operation and maintenance service provider for the overall operation and maintenance requirements, including the allocation of salaries, bonuses, benefits, etc. of the operation and maintenance department heads, project management office (PMO) personnel, organizational-level quality assurance personnel, etc.

[0043] Indirect non-labor costs refer to the allocation of non-labor costs that are not generated for a specific project by the operation and maintenance service provider but serve the overall operation and maintenance activities. It includes the rent, water and electricity, property management of the office space, the allocation of daily office expenses, and the lease, maintenance, and depreciation allocation of various office equipment.

[0044] Specifically, it is divided into basic environment operation and maintenance costs, hardware operation and maintenance costs, software operation and maintenance costs, etc.

[0045] Basic environment operation and maintenance costs : Basic environment operation and maintenance services are the operation and maintenance of infrastructure such as electricity, air conditioning, fire protection, and security that are necessary to ensure the normal operation of information systems, including routine inspections, status monitoring, response support, fault handling, performance optimization, etc. of the power supply, fire protection, security, etc. systems in the computer room. The objects of basic environment operation and maintenance services include power supply and distribution systems, generator systems, precision air conditioning systems, fresh air systems, lightning protection and grounding systems, fire protection systems, video surveillance systems, access control systems, etc. The measurement of basic environment operation and maintenance costs is to calculate the costs of providing basic environment operation and maintenance services.

[0046] ; is the direct labor cost of basic environment operation and maintenance services, is the workload of the bth basic environment operation and maintenance service, is the unit price of the bth basic environment operation and maintenance service, is the price adjustment factor of basic environment operation and maintenance services, is the direct non-labor cost of basic environment operation and maintenance services, is the indirect labor cost of basic environment operation and maintenance services, is the indirect non-labor cost of basic environment operation and maintenance services, and B is the number of types of basic environment operation and maintenance services.

[0047] The workload of basic environment operation and maintenance services is the sum of the product of the quantity of the b-th type of equipment in the basic environment, the unit workload of the b-th type of equipment, and the workload adjustment factor of the b-th type of equipment.

[0048] The unit workload is the number of man-days required for the operation and maintenance of a certain type of hardware equipment within a certain period. The workload of operation and maintenance service content is divided into the following four categories: Routine operation: the workload required for a single routine operation service of a single device × the number of times; Response support: the workload required for a single response support service of a single device × the number of times; Optimization and improvement: the workload required for a single optimization and improvement service of a single device × the number of times; Investigation and evaluation: the workload required for a single investigation and evaluation service of a single device × the number of times.

[0049] When calculating the overall workload, add up the total workloads of the above four types of operation and maintenance service content, and corresponding tailoring of operation and maintenance service content can be carried out according to the service catalog.

[0050] Hardware operation and maintenance cost : Hardware operation and maintenance services refer to services such as routine inspections, status monitoring, response support, fault handling, and performance optimization of hardware equipment (networks, hosts, storage, desktop devices, and other related devices, etc.). The objects of hardware operation and maintenance include: networks and network devices, host devices (personal computer servers, small computers, and mainframes, etc.), storage devices, desktop and peripheral devices (fixed computing terminals, mobile computing terminals, peripheral input and output devices, peripheral storage devices, and peripheral communication devices), and other hardware; the measurement of hardware operation and maintenance cost is to calculate the cost of providing hardware operation and maintenance services.

[0051] ; is the direct labor cost of hardware operation and maintenance, is the workload of the d-th type of hardware operation and maintenance service, is the unit price of the d-th type of hardware operation and maintenance service, is the price adjustment factor of hardware operation and maintenance service, is the direct non-labor cost of hardware operation and maintenance service, is the indirect labor cost of hardware operation and maintenance service, is the indirect non-labor cost of hardware operation and maintenance service, and D is the number of types of hardware operation and maintenance services; The workload of hardware operation and maintenance services is the sum of the product of the quantity of the d-th type of hardware equipment, the unit workload of the d-th type of hardware equipment, and the equipment workload adjustment factor of the d-th type of hardware equipment.

[0052] Software operation and maintenance cost : Software operation and maintenance refers to the comprehensive services provided for the operating environment of the information systems and business systems used by the demander by means of information technology and methods according to the service level requirements proposed by the demander. The service content of software operation and maintenance includes the function modification and improvement, performance tuning of software (including basic software, supporting software, application software, etc.), as well as conventional routine inspections, status monitoring, response support and other services. Software operation and maintenance cost measurement is to calculate the costs of providing these service contents.

[0053] ; is the workload of the e-th type of software operation and maintenance, is the average labor cost rate of the e-th type of software operation and maintenance, is the direct non-labor cost of software operation and maintenance, and E is the number of types of software operation and maintenance.

[0054] If the unit price of the operation and maintenance function point has been determined, the software operation and maintenance cost can be calculated according to the unit price of the operation and maintenance function point, using the following formula: Software operation and maintenance service cost = (software scale × unit price of operation and maintenance function point) × adjustment factor for operation and maintenance level requirements × adjustment factor for operation and maintenance ability × adjustment factor for operation and maintenance system and business characteristics + direct non-labor cost.

[0055] Software operation and maintenance service workload = (software scale × productivity) × adjustment factor for operation and maintenance level requirements × adjustment factor for operation and maintenance ability × adjustment factor for operation and maintenance system and business characteristics.

[0056] An electronic device, comprising: a processor; and a memory, in which computer program instructions are stored, and when the computer program instructions are run by the processor, the processor is caused to execute the dynamic scheduling method for IT device repair tasks with integrated fault prediction as described above.

[0057] A computer-readable storage medium for storing a program, which when executed by a processor implements the dynamic scheduling method for IT device repair tasks with integrated fault prediction as described above.

[0058] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0059] The present invention is described with reference to the flowcharts and / or block diagrams of systems, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate a means for implementing the functions specified in the Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.

[0060] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including an instruction means that implements the functions specified in the Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.

[0061] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operational steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in the Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.

[0062] Although the preferred embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications once they learn the basic creative concepts. Therefore, the appended claims are intended to be construed to include the preferred embodiments and all changes and modifications falling within the scope of the present invention.

[0063] Obviously, those skilled in the art can make various changes and variations to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these changes and variations.

Claims

1. A dynamic scheduling method for IT device maintenance tasks integrating fault prediction, characterized in that It includes the following steps: Construct the spatio-temporal feature matrix of the device state and the historical failure frequency matrix of IT devices, and combine the working condition correction factors stored in the database to determine the occurrence frequency matrix of each failure type of IT devices; Determine the amount of fault repair tasks based on the number of IT devices and the occurrence frequency matrix of each fault type; Determine the expected completion status of the fault repair task based on the current working status of the on-site maintenance personnel, and adjust the working mode and personnel of the on-site maintenance personnel based on the expected completion status of the fault repair task.

2. The dynamic scheduling method for IT device maintenance tasks integrating fault prediction according to claim 1, characterized in that Construct the spatio-temporal feature matrix of the device state of IT devices, including the following steps: Perform device ID-timestamp correlation cleaning on historical fault data, extract the fault component codes and severity levels from the maintenance logs through regular expressions, and generate structured fault feature vectors; Calibrate the UTC time of the device operation logs, use the dynamic time warping algorithm to align the load curves with different sampling frequencies, calculate the load change rate per hour through the first derivative, and generate the operation state time series matrix; Construct a joint interpolation space for the data of the temperature and humidity sensors and the vibration spectrum analyzer, generate a minute-level synchronized data stream through cubic spline interpolation, and calculate the temperature-vibration covariance coefficient per device per hour to form the environmental impact tensor; Obtain the spatio-temporal feature matrix of the device state through spatio-temporal tensor product operation on the structured fault feature vectors, the operation state time series matrix, and the environmental impact tensor; Construct the historical fault frequency matrix, including the following steps: Obtain the device historical fault data from the database, group by device model, calculate the average annual incidence rate of each fault type, and construct it into the historical fault frequency matrix.

3. The dynamic scheduling method for IT device maintenance tasks integrating fault prediction according to claim 2, characterized in that Determine the occurrence frequency matrix of each fault type of IT devices, including the following steps: Obtain the trained multi-task prediction model stored in the database. The multi-task prediction model adopts a dual-channel gated recurrent unit network, where the main channel input is the device status spatio-temporal feature matrix , and the auxiliary channel input is the historical failure frequency matrix and the working condition correction factor ; Use the modified Poisson regression to output the occurrence frequency of each fault type of each IT device within the future time window; ； ； ； Among them, is the gating signal, is the gating weight matrix, is the Sigmoid function, is the calculation result of the fused feature, is the element-wise product, is the bias term for the c-th fault type, is the sensitivity weight of the c-th fault type of the i-th IT device to the m-th dimensional fused feature, where M is the total number of dimensions of the fused feature, is the occurrence frequency of the c-th fault type of the i-th IT device within the future time window; Dynamically adjust the predicted occurrence frequency according to the real-time data flow, and construct it into the occurrence frequency matrix of each fault type of IT devices: , where is the occurrence frequency of dynamic adjustment prediction, representing the frequency of the c-th type of failure occurring in the i-th type of IT device within the future time window, is the forgetting factor, is the sliding window length, is the actual number of secondary failures occurring within the sliding window length.

4. The dynamic scheduling method for IT device maintenance tasks integrating fault prediction according to claim 3, characterized in that, Determine the amount of fault repair tasks based on the number of IT devices and the occurrence frequency matrix of each fault type. The process is as follows: , where is the amount of fault repair tasks, is the quantity of the i-th type of IT equipment.

5. The dynamic scheduling method for IT device maintenance tasks integrating fault prediction according to claim 1, characterized in that Determine the expected completion status of the fault repair task based on the current working status of the on-site maintenance personnel, including the following steps: Obtain the maintenance work efficiency of each on-site maintenance personnel for each fault type of each IT device, and calculate the average fault maintenance efficiency of each on-site maintenance personnel; Obtain the on-site time of each current on-site maintenance personnel, and calculate the manual maintenance workload: , where is the workload of manual maintenance, is the average fault maintenance efficiency of the a-th on-site maintenance personnel, is the on-site time of the a-th on-site maintenance personnel, and A is the number of current on-site maintenance personnel; Subtract the manual maintenance workload from the breakdown maintenance task volume to obtain the difference If the difference is less than 0, the expected completion status of the fault repair task is unable to be completed; If the difference is not less than 0 and not greater than the set threshold, the expected completion status of the fault repair task is able to be completed; If the difference is greater than the set threshold, the expected completion status of the fault repair task is that there are too many on-site personnel.

6. The dynamic scheduling method for IT device maintenance tasks integrating fault prediction according to claim 5, characterized in that, Adjust the working mode and personnel of the on-site maintenance personnel based on the expected completion status of the fault repair task, including the following steps: If the expected completion status of the fault repair task is unable to be completed, conduct internal transfer of maintenance personnel; If the expected completion status of the fault repair task is able to be completed, maintain the current status of the on-site maintenance personnel; If the expected completion status of the fault repair task is that there are too many on-site personnel, conduct external transfer of maintenance personnel or change the working mode of the maintenance personnel.

7. The dynamic scheduling method for IT device maintenance tasks integrating fault prediction according to claim 6, characterized in that, Conduct internal transfer of maintenance personnel, including the following steps: Determine the list of the most suitable personnel from the list of schedulable personnel based on the adaptability calculation model; Based on the constraint conditions, screen the personnel in the list of the most suitable personnel to determine the best combination of internal transfer personnel.

8. The dynamic scheduling method for IT device maintenance tasks integrating fault prediction according to claim 7, characterized in that Determine the list of the most suitable personnel from the list of schedulable personnel based on the adaptability calculation model, including the following steps: Determine the work characteristic matrix of each schedulable personnel in the list of schedulable personnel, including the set of historical maintenance fault types, the set of the number of times of completing maintenance for each fault type, the set of the duration of completing maintenance for each fault type, and the average work efficiency; Input the work characteristic matrix of each schedulable personnel into the adaptability calculation model to obtain the maintenance adaptability of each schedulable personnel; Arrange the maintenance adaptabilities in descending order, and obtain the top R schedulable personnel as the list of the most suitable personnel.

9. The dynamic scheduling method for IT device maintenance tasks integrating fault prediction according to claim 8, characterized in that, The adaptability calculation model is expressed as: ； Among them, For maintaining the adaptability, For the average working efficiency, Is the set of historical maintenance fault types, Is the set of the number of times of completing maintenance for each fault type, Is the set of the duration of completing maintenance for each fault type, Is the set of required historical maintenance fault types, Is the set of the number of times of completing maintenance for each required fault type, Is the set of the duration of completing maintenance for each required fault type, Is the cosine similarity function.

10. The dynamic scheduling method for IT device maintenance tasks integrating fault prediction according to claim 7, characterized in that, Based on the constraint conditions, screen the personnel in the list of the most suitable personnel to determine the best combination of internal transfer personnel, including the following steps: Randomly combine the original on-site personnel with the personnel in the list of the most suitable personnel to obtain a combination of maintenance personnel, where the combination of maintenance personnel includes all the original on-site personnel and at least one person in the list of the most suitable personnel; Evaluate each combination of maintenance personnel based on the constraint conditions to determine the best combination of maintenance personnel as the best combination of internal transfer personnel, where the constraint conditions are: ； Among them, is the amount of fault repair tasks, is the average work efficiency of the p-th maintenance personnel in the maintenance personnel combination, is the on-site duration of the p-th maintenance personnel in the maintenance personnel combination, is the redundancy factor, is the operation and maintenance cost function.

Images